The paradox of creativity in generative AI: high performance, human-like bias, and limited differential evaluation

Creativity plays a crucial role in helping individuals and organisations generate innovative solutions to arising challenges. To support this creative process, generative Artificial Intelligence (AI), such as ChatGPT is being used increasingly. However, whether such a generative AI model can truly enhance creativity or whether it exhibits similar creative biases to humans is unclear. This study, conducted in 2025, consisted of an experiment which involved ChatGPT-4o performing the egg task, a creativity task which measures fixation bias and original idea generation (expansion). The AI model's results were compared both to a sample of 47 human participants and to aggregated data from eight previous studies using the same procedure with the egg task. This dual comparison provides a comprehensive perspective on creative biases in both AI and humans at multiple levels. While ChatGPT demonstrated greater productivity than humans, it exhibited a comparable fixation bias, with most ideas falling within conventional categories. Furthermore, the model showed a limited capability to differentially evaluate originality, as it struggled to distinguish between original and conventional ideas, unlike humans who are typically able to make this distinction. In conclusion, although generative AI demonstrates impressive fluency by producing a large number of creative ideas, its inability to critically assess their originality and overcome the fixation bias highlights the necessity of human involvement, particularly for properly evaluating and filtering the ideas generated.

Joy Desdevises